I, robot – Can you translate without understanding the meaning?
Abstract
In this thought-provoking article, published in the ITI Bulletin in May 2024, Michael Farrell explores whether understanding is truly necessary for accurate translation. Through two carefully designed linguistic puzzles, he demonstrates how machines, unlike humans, rely purely on pattern recognition, statistical inference and token embeddings to translate between languages without grasping meaning. Drawing on natural language processing concepts and the principle that “you shall know a word by the company it keeps,” Farrell highlights both the power and the limitations of machine translation and generative AI. The article sets the stage for a follow-up discussion on why human insight remains essential in the translation process, which appears in the following edition of the ITI Bulletin in July 2024.
Published in
ITI Bulletin, May-June 2024
Engaging copy translated literally into English, without taking account of differences in linguistic, semantic and cultural expressions, at best leaves much to be desired and at worst provokes hysterical laughter.
Thanks to my scientific background, I specialize in technical translations. Over the years I have acquired experience in transcreating advertising copy and press releases primarily for the promotion of technology products.
All clients of mine who have sent me urgent requests are satisfied with my response. 
